pine-scripts

Support & Resistance Zones

A detector-driven support/resistance zone engine. Independent detectors (pivot clusters, multiple tests, base/supply-demand, order blocks, fair value gaps, price gaps, equal-high/low liquidity) each emit ZoneCandidate evidence into a central registry, which is the only part of the script allowed to create, merge, score or draw a zone. Detectors never draw directly — detection and rendering are deliberately separated so the registry can be validated, reused by other WavesUnchained indicators, and eventually extracted into a shared library.

Architecture

Detectors -> ZoneCandidate -> candidate validation -> Zone registry ->
clustering/merging -> confluence -> scoring -> lifecycle -> renderer + logs + statistics
  • Detectors only produce ZoneCandidate objects (direction, price range, origin bar, source code, initial score, reason string). They hold no drawing logic.
  • Registry (f_ingestCandidate) validates a candidate’s width against ATR, then either merges it into the nearest valid overlapping/nearby same-direction zone or creates a new Zone with a stable id.
  • Confluence (f_applyConfluence) adds score once per source (EMA, VWAP, Fibonacci, psychological level, relative volume) to a zone at the moment it is touched, independent of which detector created it — not continuously, so a wide zone can’t accumulate confluence merely by being wide.
  • Lifecycle mutates only on confirmed bars, advances each zone through age, break, touch and reaction checks, and re-checks each live zone’s width against Maximum Merged Zone Width · ATR so a zone merged during a high-volatility period doesn’t stay frozen that wide once volatility drops. A new zone and a break/flip bar cannot validate themselves as a retest.
  • Renderer only draws the Maximum Visible Zones Per Side zones nearest to price that clear Minimum Visible Zone Score, are on their actual side of price (support at/below, resistance at/above), and don’t overlap a selected zone of the same direction.

Features

  • Pivot cluster (PC) detector from ATR-tolerant clustering of confirmed pivots retained for the configured history window; MT is normally earned through later real zone tests and falls back to stricter pivot clustering only if PC was not accepted
  • Base / Supply & Demand detector — compact range followed by a displacement exit; the shared raw pattern is one source (SD when enabled, otherwise BASE), never BASE+SD double evidence
  • Order Block (OB) detector — displacement + prior-structure break + last opposite-color origin candle
  • Fair Value Gap (FVG) and residual unfilled classical Price Gap (GAP) detectors, tracked as separate sources
  • Equal High / Equal Low Liquidity (LIQ) detector from clusters of same-level confirmed pivots in retained history
  • Central zone registry with ATR-based merge distance, merged-width cap and stable Zxxx IDs
  • Confluence scoring: EMA, VWAP, Fibonacci (rolling high/low), psychological price step, relative volume
  • Lifecycle: touch detection with cooldown, reaction validation against a configurable ATR target and horizon, confirmed break detection, optional Support <-> Resistance flip on break (BO source)
  • Nearest-to-price renderer — top N zones per side, score-filtered, non-overlapping, on-chart state label, with hover tooltips (state, sources, score, touches, reactions, breaks, origin/created bar)
  • Pine Log event logging (CREATE/MERGE/REJECT/TOUCH/REACTION/WEAK/BREAK/FLIP) and an optional validation statistics table for Validation/Debug operating modes

Scoring

score (0-100, clamped) is a heuristic confluence/quality score, not a statistical probability. It accumulates from:

  • The initiating detector’s configured weight (Pivot Cluster, Multiple Tests, Base, Supply/Demand, Order Block, Fair Value Gap, Price Gap, Liquidity)
  • Each additional detector whose evidence merges into the same zone (once per source code, tracked via the sources string, e.g. PC+OB+FIB)
  • Confluence hits (EMA/VWAP/Fib/Psychological/Volume), each awarded once per zone
  • A Breakout/Flip bonus (BO) the first time a zone is confirmed broken and flips direction

All weights live in the 10 · Scoring input group so relative detector/confluence importance can be tuned without touching code.

Lifecycle states

NEW -> TESTED -> CONFIRMED on a successful reaction, or -> WEAK if the reaction horizon expires without one. A confirmed break either sets BROKEN (kept for Keep Broken Zones bars, then INVALID) or, if Convert Broken Zones Into Flip Zones is on, sets FLIPPED and reverses dir. INVALID zones are functionally dead but remain in storage until pruned by Maximum Stored Zones. With Show Zone State on (default), the current state is appended to the on-chart label, e.g. OB+FIB+MA · 28 · TESTED — validity is visible without hovering.

Zone age vs. provenance

The rendered box’s left edge is anchored to createdBar (when the Zone entity itself was instantiated) rather than originBar (the earliest evidence ever merged into it). originBar only ever regresses further into the past as a zone absorbs more evidence over time — anchoring the box to it would make every long-lived, frequently-reinforced zone look equally ancient regardless of how recently it was actually reinforced. originBar remains visible in the hover tooltip as “how far back does evidence for this price area go,” a genuinely different and still useful question from “how long has this exact zone existed.”

Zone codes

The label on each rendered zone is its sources string with | replaced by + (e.g. PC+OB+FIB), optionally followed by · <score>. Codes: PC pivot cluster, MT multiple tests, BASE/SD base / supply-demand, OB order block, FVG fair value gap, GAP price gap, LIQ liquidity, BO breakout/flip, MA/VWAP/FIB/PSY/VOL confluence.

Status

This is a 0.1.11 engine-foundation build: the architecture, detectors, merging, lifecycle, confluence, scoring, renderer, logging and validation table exist end-to-end, but no detector threshold or scoring weight has been calibrated against real charts, and the script has not yet been run through the TradingView Pine Editor compiler. The v0.1.6 static-review fixes ensure that lifecycle events occur only after a confirmed bar, fresh zones do not validate themselves, GAP boxes are residual unfilled gaps, reinforcement extends analysis lifetime, and the renderer honours actual price sides. v0.1.7 makes equal-high/low detection robust to intervening pivots, v0.1.8 removes duplicate BASE/SD evidence, v0.1.9 chooses the nearest valid merge target, v0.1.10 retains the configured pivot-history window, and v0.1.11 removes routine PC/MT double scoring. See CHANGELOG.md and todo.md; the next priority remains compiling and isolating the registry before trusting the speculative detectors (OB, SD, LIQ).

Known limitation — width re-check can cut off an in-progress touch

The v0.1.1 width re-check re-centers an over-wide zone around its midpoint every bar, including bars where price is actively inside it. If a zone was wide because that width was itself meaningful evidence (e.g. a genuinely broad liquidity pool), shrinking it mid-touch could clip the interaction. Not yet observed on a real chart — flagged here so it’s the first thing to check if reaction/touch counts look off after this change.

Edit this page on GitHub